The AI Assistant configures your test for you.
When the recorded replay throws a runtime error, the AI Assistant examines the HTTP traffic around the failure, identifies the pattern, and writes the detection rule. Behind it sits Automatic State Management (ASM), the rules-based expert system Web Performance has refined since the late 1990s. The AI catches the long tail; ASM handles the well-known patterns up-front.
What the AI Assistant actually does.
Four steps, every time a replay fails on a missing or stale dynamic value:
1. Error detected
A replay fails because a session token, cookie, or correlated value is missing or stale.
2. AI analyzes
The assistant examines the HTTP traffic around the failure, identifies the pattern in the prior response, and determines the extraction strategy.
3. Rule generated
The assistant writes the detection algorithm and explains what it is about to do before applying it. You accept, correct, or override.
4. Replay succeeds
The rule applies, the test runs, and the configuration persists for every future run of the same scenario.
Specifics from recent engagements: an OAuth login flow that previously cost 2 hours of manual correlation typically configures in around 4 minutes. A 50-page checkout scenario that took a full day to script in legacy tools typically configures in around 8 minutes. Across the board, the AI Assistant compresses configuration roughly 6x compared to manual work.
Underneath: a 27-year-old expert system.
The AI Assistant is the orchestrator. The correlation precision comes from Automatic State Management, the rules-based expert system Web Performance has refined since 1999 across professional performance engagements. ASM scans every recorded transaction up-front and configures the well-known dynamic-value patterns automatically:
- Cookies
- Form fields (when not user-entered)
- Hidden form fields, including .NET VIEWSTATE and related fields
- Query parameters
- ETag and If-Modified-Since headers
- Authentication headers
- Proxy headers
- XML fields
- JSON fields
This is the "boring" part of the work. The AI Assistant only has to think about cases ASM cannot resolve from its scan: proprietary patterns, unusual framework quirks, or fields that only surface as runtime failures during replay.
OAuth2 and modern web frameworks.
ASM detects OAuth2 Bearer tokens, refresh tokens, and JWT in JSON responses and correlates them with Authorization headers in subsequent requests. No manual configuration. Beyond OAuth2, the modern-framework patterns ship as built-in rules:
- React and Next.js · CSRF tokens in meta tags, state-management tokens
- Angular · XSRF tokens, JWT authentication
- Vue.js · CSRF protection, Vuex state tokens
- Node.js / Express · session middleware, csurf tokens
- GraphQL APIs · operation IDs, authentication tokens
- REST APIs · OAuth2, API keys, request signatures
Custom rules for the edge cases.
For proprietary patterns, custom authentication schemes, or any framework ASM does not yet ship a built-in rule for, custom detection rules use plain .properties configuration files. Define a regex extraction pattern, scope it to a request or response context, and ASM picks it up at runtime.
- No code changes, no recompile, no restart
- Field-deployable: drop the file in and the rule is live
- Framework-agnostic: works for any proprietary API or custom auth scheme
- Composable with built-in ASM rules and with the AI Assistant
One customer testing a custom Java framework added OAuth2 token detection for a proprietary API in under five minutes using a simple configuration file, no source-code access needed.
The end-to-end flow.
- Record. Capture HTTP traffic from a real browser session against your application.
- Detect. ASM scans the captured traffic and identifies the well-known dynamic-value patterns.
- Extract. ASM and the AI Assistant generate detection rules that pull dynamic values from responses.
- Replay. The extracted values apply during load-test playback. Runtime errors trigger the AI Assistant to generate new rules as needed.
No scripting language. No manual correlation. The configuration that took hours in script-based load testing tools takes minutes here.
"The design time is split by 4 or 5, even more if scenarios are complex."
That quote is from Jean-Luc ORTS, the project manager who led load testing at Caisse Nationale d'Assurance Vieillesse (the French equivalent of the Social Security Administration), explaining why CNAV switched to WPLoadTester from a script-based tool. The 4x-to-5x figure he gave reflects ASM alone, before the AI Assistant existed. Recent engagements with the AI Assistant added on top run closer to 6x.
The numbers are real engagements. The methodology has been refined across 27 years of professional performance work. The AI Assistant follows that methodology rather than reinventing it.
Try the AI Assistant on your application.
The AI Assistant and ASM ship together in WPLoadTester 7. Request the beta to run them on your test scenarios, or download the free single-machine edition to evaluate locally.
Comparing tiers? See the Free vs Pro split.